ep-engine mangles last entry in failover log on unclean shutdown even when everything is safely on disk

This is what I get after saving a number of docs into vbucket other and successfully waiting for them to hit disk.

Before shutdown:

root@beta:~/src/altoros/moxi/repo30# rlwrap ../send-diag-eval lh:9000
xdcr_upr_streamer:get_failover_log("other", 7).
RestClient.post "lh:9000/diag/eval", "xdcr_upr_streamer:get_failover_log(\"other\", 7).\n", "Accept"=>"/; q=0.5, application/xml", "Accept-Encoding"=>"gzip, deflate", "Content-Length"=>"48"

1. => 200 OK | 76 bytes:[ {13226830787664969728,0},{2722187756529715712,48},{4817340327337429639,96}]
   
   
   [{13226830787664969728,0}

   ,

   {2722187756529715712,48},{4817340327337429639,96}]
   
   
   After Ctrl-C (unclean shutdown) and restart:
   
   root@beta:~/src/altoros/moxi/repo30# rlwrap ../send-diag-eval lh:9000
   xdcr_upr_streamer:get_failover_log("other", 7).
   RestClient.post "lh:9000/diag/eval", "xdcr_upr_streamer:get_failover_log(\"other\", 7).\n", "Accept"=>"/; q=0.5, application/xml", "Accept-Encoding"=>"gzip, deflate", "Content-Length"=>"48"
   # => 200 OK | 76 bytes:[{13226830787664969728,0},{2722187756529715712,48}

   ,

   {4817340327337430016,96}]
   
   
   [{13226830787664969728,0},{2722187756529715712,48},{4817340327337430016,96}

   ]

Note how last entry uuid is changed.

Currently that's causing my xdcr work in progress checkpointing to fail to resume from xdcr checkpoint without good reason.